MetHis

MetHis reconstructs complex two-way genetic admixture histories by forward-time simulation of independent single nucleotide polymorphisms (SNPs) and integration with Approximate Bayesian Computations (ABC) frameworks.

Key Features:

• Forward-time simulation: Simulates independent single nucleotide polymorphisms (SNPs) within a two-way admixed population.
• Admixture scenario models: Supports multiple admixture pulses and monotonically decreasing or increasing admixture contributions at each generation.
• Prior distributions: Allows model-parameter values to be defined through prior distributions.
• Summary statistics: Computes 24 summary statistics describing genetic diversity and the moments of individual admixture fractions.
• Approximate Bayesian Computations (ABC) integration: Employs Random-Forest ABC for scenario choice followed by Neural-Network ABC for posterior parameter estimation.
• Complement to likelihood methods: Applicable when maximum-likelihood approaches based on allele frequencies and linkage disequilibrium (LD) from dense genome-wide datasets are inadequate for complex multi-pulse admixture.

Scientific Applications:

• African American and Barbadian admixture reconstruction: Applied to African American and Barbadian populations, finding monotonically decreasing European and African contributions over time better explained the observed genetic data than multiple admixture pulses.
• Introgression decay analysis: Identified distinct trajectories of introgression decay between the two studied populations.
• Cross-population/species admixture studies: Reconstructs detailed admixture histories across various populations and species when forward simulation and ABC are required.

Methodology:

Performs forward-time simulation of independent SNPs in a two-way admixed population under scenarios including multiple pulses or monotonically changing contributions, samples model parameters from prior distributions, computes 24 summary statistics describing genetic diversity and moments of individual admixture fractions, and applies Random-Forest ABC for scenario choice followed by Neural-Network ABC for posterior parameter estimation.